R. Mullick, H. T. Nguyen, Y. P. Wang, J. K. Raphel, and R. Raghavan
Center for Information-enhanced Medicine
|
ABSTRACT |
The Center for Information-enhanced Medicine (CIeMed) has
been
actively involved in the analysis, visualization, and
development of
applications based on the Visible Human(TM) dataset. In this
paper we
summarize our efforts to explore this vast dataset. Our
projects
include complete segmentation and labeling, direct volume
rendering
for the entire RGB-volume, real-time catheter simulation
using the
dataset as a virtual patient, and 2D/3D network based anatomy
atlases.
|
HUMAN
|
The human anatomy visualization project within CIeMed has
segmented
and labeled the Visible Human(TM)(VH) male dataset into over
300
tissues and organs. This task has been achieved by applying
semi-automatic and manual volume segmentation methods to the
VH
cyro-section data and the registered CT data. An effort to
depict
these labeled regions, using color, is presented in Figure 1(a).
|
VHD Female Head VHD Female Heart VHD Male Thorax VHD Male Pelvis |
Figure 1: (a) Segmented Visible Human male Data; (b)
Full body translucent
RGB
|
Furthermore, we have developed a methodology to RGB volume
render the
entire photo volume using only limited memory systems. This
approach,
Partitioned Volume Rendering [1
-2] based on a divide-and-conquer
algorithm, allows selective translucent visualization of
various
tissues, organs and systems of the human anatomy and their
inter-relationships [Figure 1(b)]. An
interactive software environment [Figure
1(c)] allows the user to define the numerous parameters
for
visualization and to create an animation of the same by
varying these
parameters using key-frames. CIeMed has also developed
Vortex, an
interactive volume rendering approach using hardware-based 3D
texture
mapping. Vortex offers real-time translucent viewing [Figure 2(a)] and perspective fly-through
capabilities. Using these technologies, CIeMed and Alex Tsiaras with Time-Warner, are producing a comprehensive educational study of the human body, understandable and appealing to the lay public and the medical community.
|
daVINCI:
|
daVinci is a real-time simulation prototype for vascular catheterization procedures. It allows interventional radiologists and students to learn and rehearse difficult navigational tasks for intravascular procedures. daVinci also allows device manufacturers to design and test catheter and guidewire shapes and specify their physical properties. |
Cath. Navigation |
CIeMed has used the Visible Human dataset as a virtual
patient for the
navigational and visual needs of this simulator. The primary
arterial vasculature was extracted from the VH photo data.
This
primary structure served as a roadmap for extending the
arterial
model by addition of secondary and tertiary networks from
other
scanned data. A physical model based on this vasculature is
used
for real-time computation in daVinci [3].
The final vasculature
[Figure 2(b)] and the intravascular
devices
(catheter, guidewire, etc.) are registered to x-ray
projection
views of the Visible Human CT volume to offer realistic
fluoroscopic images
and controls (rotate, translate, magnify) as observed in a
Catheterization Laboratory.
|
Figure 2: (a) Interactive volume rendering using 3D
texture mapping; and (b)
Fluoroscopic view (using CT data) with vasculature and
catheter
highlighted.
|
INTERACTIVE
|
The Interactive Internet Atlases, an extension of [4], are based on a Producer Consumer Observer model with the atlas producer generating the data set, atlas consumer providing the rendering and atlas observer capturing user's interactions. |
Figure 3: (a) Java Applet to view Labeled VH Data. (b)
Interactive 3D viewing of
the VH male cortex over Internet using VRML and SGI Webspace.
|
Various architectures which
trade off the communication, computation and interactivity
are
possible within this model. For example, applications such as
interactive 2D navigation of the atlas data, the atlas
producer acts
as a data server for the atlas requests generated by the
client
applets [Figure 3(a)]. In situations
where
3D rendering capabilities are required, the atlas producer is
required
to generate data sets corresponding to the given orientation.
This
results in increased communication costs and require
specialized
servers to handle multiple rendering requests from clients.
An
alternative is to equip the applets with the ability to
handle VRML
and retain the architecture similar to the 2D atlas [Figure 3(b)].
|
ACKNOWLEDGMENTS |
The authors wish to thank Solaiyappan Meiyappan for the
development of
Vortex, Ng Yew Choong for coding the Java applets, Ms. Jin
Xiaoyang
and all the students from the National University of
Singapore
involved in the Human Anatomy Visualization project for their
role in
segmenting the dataset. We would also like to thank our
collaborators
Alexander Tsiaras, Anatomical Travelogue, and Dr. Jim
Anderson, Johns
Hopkins University.
|